Microfluidic and SPR battery water state identification method, equipment and medium

Abstract

The invention discloses a microfluidic and SPR battery water state identification method, equipment and a medium, and relates to the technical field of fuel cells. The method comprises the steps that the hollow optical fiber is embedded into a bipolar plate runner of the fuel cell, a cavity of the hollow optical fiber is a micro-fluidic channel, and the micro-fluidic channel is used for inflow and outflow of liquid in the fuel cell; transmitting the optical signal to the hollow optical fiber, and coating the inner wall of the cavity with a metal layer which is used for exciting SPR (Surface Plasmon Resonance); collecting a spectral signal after interaction of the liquid and the optical signal in real time through a microfluidic channel, and generating a two-dimensional spectrogram; carrying out dimensionality reduction on the two-dimensional spectrogram, extracting gray level image features, and generating a one-dimensional feature vector; and inputting the one-dimensional feature vector into a pre-trained classifier, classifying the state of the fuel cell through supervised and unsupervised mixed learning, and outputting a diagnosis result. According to the method, the state of the liquid in the battery can be accurately diagnosed in real time.

Description

Technical Field

[0001] The present application relates to the field of fuel cell technology, and in particular to a method, device, and medium for identifying the water status of a battery using microfluidics and SPR. Background Art

[0002] Against the backdrop of profound changes in the global energy landscape and increasingly stringent environmental standards, fuel cells, as a new energy technology with enormous potential, are gradually finding widespread application in the energy sector. Fuel cells are devices that, based on electrochemical reactions, directly and efficiently convert the chemical energy of a fuel and an oxidant into electrical energy. Compared to traditional energy sources, fuel cells offer significant advantages in reducing environmental pollution and reliance on non-renewable resources like oil, providing strong technical support for achieving sustainable energy development.

[0003] However, fuel cells still face numerous technical challenges in their journey towards large-scale commercial application. During fuel cell operation, the electrochemical reaction between hydrogen and oxygen produces a large amount of water molecules at the cathode. These water molecules are transported within the fuel cell and eventually discharged. However, excessive or insufficient water content in a fuel cell can affect its performance and lifespan. Therefore, proper water management is crucial to ensuring the proper operation and extended lifespan of the fuel cell.

[0004] Through the above analysis, the problems and defects of the existing technology are as follows: The solid optical fiber in the existing technology has a limited contact area with the liquid, and is easily destroyed in the liquid flow distribution after being implanted in the flow channel, causing large interference to the liquid flow. In addition, the integration of microfluidic technology and SPR is insufficient, the monitoring sensitivity is low, and the battery status cannot be identified in time. Summary of the Invention

[0005] The embodiments of the present application provide a battery water status identification method, device and medium using microfluidics and SPR, which can solve the problems in the prior art of limited contact area between solid optical fiber and liquid, easy destruction of liquid flow distribution after implantation into the flow channel, large liquid flow interference, insufficient integration of microfluidics technology and SPR, and low monitoring sensitivity.

[0006] In the first aspect, an embodiment of the present application provides a microfluidic and SPR battery water status identification method, the method comprising: embedding a hollow optical fiber into a bipolar plate flow channel of a fuel cell, wherein the cavity of the hollow optical fiber is a microfluidic channel, the microfluidic channel is used for the inflow and outflow of liquid inside the fuel cell, and the hollow optical fiber is semi-open; transmitting an optical signal to the hollow optical fiber, coating a metal layer on the inner wall of the cavity, and the metal layer is used to excite SPR; through the microfluidic channel, collecting spectral signals after the interaction between the liquid and the optical signal in real time to generate a two-dimensional spectrum graph, the two-dimensional spectrum graph including refractive index, light intensity and resonance wavelength; performing dimensionality reduction on the two-dimensional spectrum graph, and extracting grayscale image features to generate a one-dimensional feature vector; inputting the one-dimensional feature vector into a pre-trained classifier, classifying the state of the fuel cell through supervised and unsupervised hybrid learning, and outputting a diagnosis result, the diagnosis result including normal state, flooded state or membrane dry state.

[0007] In one implementation of the present application, a metal layer is coated on the inner wall of the cavity, and the metal layer is used to excite SPR, specifically including: coupling the optical signal to the hollow fiber through a multimode optical fiber; depositing an adhesion layer on the cavity surface of the hollow fiber, and the material of the adhesion layer is chromium or titanium; coating a metal film on the adhesion layer, and modifying nanoparticles on the surface of the metal film, and the metal film is used to excite surface plasmon polaritons, and the material of the metal film is gold or silver.

[0008] In one implementation of the present application, the method also includes: when coupling the optical signal through a multimode optical fiber, emitting optical signals of different wavelengths in sequence to obtain the resonance angle offset and the light intensity change, and cross-validating them with the mapping relationship. The mapping relationship is obtained by adding interference during the experimental phase to obtain the mapping relationship with the spectral signal in the normal state, flooded state and membrane dry state; when the confidence level of the diagnostic result is lower than the preset threshold for a consecutive preset number of times, switching to the redundant sensing mode; correcting the mapping relationship through a self-calibration algorithm, and incorporating the fault false alarm rate into the loss function of the classifier.

[0009] In one implementation of the present application, the dimensionality of the two-dimensional spectrogram is reduced, and the grayscale image features are extracted to generate a one-dimensional feature vector, specifically including: converting the two-dimensional spectrogram into a grayscale image and performing pixel compression; applying a Gabor filter to the compressed image to extract frequency domain features, and dividing the image into blocks to calculate the local average value to generate a low-dimensional feature vector.

[0010] In one implementation of the present application, a one-dimensional feature vector is input into a pre-trained classifier, and the state of the fuel cell is classified through supervised and unsupervised hybrid learning to output a diagnostic result, specifically including: extracting a preset number of low-dimensional feature vectors for labeling, and initializing the classifier using the labeled data set; based on feature space uncertainty sampling, incrementally learning unlabeled low-dimensional feature vectors with a confidence level lower than a threshold to update the classifier, and evaluating the updated classifier.

[0011] In one implementation of the present application, the method also includes: performing plasma activation treatment on the optical fiber cavity, and setting a vortex generator at the connecting position between the microfluidic channel and the bipolar plate flow channel; using a multi-scale direction combination when extracting frequency domain features using a Gabor filter; and adjusting the temperature of the microfluidic channel to the laboratory temperature value.

[0012] In one implementation of the present application, after inputting a one-dimensional feature vector into a pre-trained classifier, classifying the state of the fuel cell through supervised and unsupervised hybrid learning, and outputting a diagnosis result, the method also includes: adjusting the liquid flow rate according to the diagnosis result to balance the water content inside the fuel cell; adjusting the liquid flow rate according to the diagnosis result to balance the water content inside the fuel cell; when the diagnosis is a flooded state, increasing the preset discharge rate of the liquid in the microfluidic channel; when the diagnosis is a membrane dry state, reducing the preset discharge rate of the liquid in the microfluidic channel.

[0013] In one implementation of the present application, after collecting the spectral signal after the interaction between the liquid and the light signal in real time through the microfluidic channel, the method also includes: independently extracting the time series characteristics of the microfluidic channel and inputting the diagnostic results into the multi-physics field coupling equation; modeling the spatial correlation between channels through the graph convolutional network, predicting the regional diffusion trend within a preset time, and generating preventive control instructions.

[0014] In a second aspect, an embodiment of the present application also provides a microfluidic and SPR battery water status identification device, the device including at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can: embed a hollow optical fiber into the bipolar plate flow channel of the fuel cell, wherein the cavity of the hollow optical fiber is a microfluidic channel, the microfluidic channel is used for the inflow and outflow of liquid inside the fuel cell, and the hollow optical fiber is semi-open; transmit the light signal to the hollow optical fiber, and coat a metal layer on the inner wall of the cavity, and the metal layer is used to excite SPR; through the microfluidic channel, the spectral signal after the interaction between the liquid and the light signal is collected in real time to generate a two-dimensional spectrum diagram, the two-dimensional spectrum diagram includes refractive index, light intensity and resonance wavelength; the two-dimensional spectrum diagram is reduced in dimension, and the grayscale image features are extracted to generate a one-dimensional feature vector; the one-dimensional feature vector is input into a pre-trained classifier, and the state of the fuel cell is classified through supervised and unsupervised hybrid learning, and a diagnosis result is output, which includes a normal state, a flooded state or a membrane dry state.

[0015] On the third aspect, the embodiment of the present application also provides a microfluidic and SPR battery water state identification non-volatile computer storage medium, which stores computer executable instructions, and the computer executable instructions are set to: embed a hollow optical fiber into the bipolar plate flow channel of the fuel cell, wherein the cavity of the hollow optical fiber is a microfluidic channel, the microfluidic channel is used for the inflow and outflow of the liquid inside the fuel cell, and the hollow optical fiber is semi-open; transmit the light signal to the hollow optical fiber, and coat the inner wall of the cavity with a metal layer, which is used to excite SPR; through the microfluidic channel, the spectral signal after the interaction between the liquid and the light signal is collected in real time to generate a two-dimensional spectrum diagram, which includes refractive index, light intensity and resonance wavelength; reduce the dimension of the two-dimensional spectrum diagram, and extract grayscale image features to generate a one-dimensional feature vector; input the one-dimensional feature vector into a pre-trained classifier, and classify the state of the fuel cell through supervised and unsupervised hybrid learning, and output a diagnosis result, which includes a normal state, a flooded state or a membrane dry state.

[0016] The embodiments of the present application provide a microfluidic and SPR battery water status identification method, device and medium. The cavity setting of the semi-open hollow optical fiber used significantly increases the contact area between the liquid and the optical fiber, thereby effectively improving the measurement accuracy and making the measurement results more reliable; the liquid between the bipolar plates can flow and transfer through the cavity, which to a certain extent reduces the interference of the presence of the optical fiber on the flow characteristics and distribution state of the liquid between the bipolar plates; the hollow optical fiber has higher real-time performance and accuracy, and is more conducive to real-time monitoring of the battery status inside the fuel cell.

[0017] Using SPR technology, the liquid state inside the battery can be detected in real time and dynamically, which is conducive to analyzing the mechanism of the water state inside the fuel cell. Compared with ordinary optical fibers, this technology also has higher sensitivity and stability, which enables it to sense slight changes in water content, thereby improving the accuracy of detection.

[0018] The combination of SPR technology and microfluidic technology improves the overall performance and reduces the use of external equipment. Compared with traditional fiber optic detection and microfluidic detection, this solution has higher sensitivity, stronger anti-electromagnetic interference ability, more flexible use, and greater equipment stability.

[0019] The machine learning model used, namely the ASSDRB classifier, can actively learn from the remaining unlabeled images in offline mode. The method continuously self-updates and evolves without the need to retrain from scratch, which improves learning efficiency and can quickly adapt to new data, which further improves the stability of the model in different environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 A flow chart of a method for identifying battery water status using microfluidics and SPR provided in an embodiment of the present application; Figure 2 A schematic diagram of the cross-sectional structure of a hollow fiber for a microfluidics and SPR battery water state identification method provided in an embodiment of the present application; Figure 3 A schematic diagram of a clover-shaped optical fiber structure for a microfluidics and SPR battery water state identification method provided in an embodiment of the present application; Figure 4 A schematic diagram of the hollow-core optical fiber distribution of a microfluidics and SPR battery water state identification method provided in an embodiment of the present application; Figure 5 A schematic diagram of the optical fiber SPR structure of a microfluidics and SPR battery water state identification method provided in an embodiment of the present application; Figure 6 A schematic diagram of the data processing flow of a microfluidics and SPR battery water state identification method provided in an embodiment of the present application; FIG7 (a) is a schematic diagram of a flow channel after installing a hollow-core optical fiber in a microfluidic and SPR battery water state identification method provided in an embodiment of the present application; FIG7 (b) is a schematic diagram of a flow channel without a hollow-core optical fiber installed in a microfluidic and SPR battery water state identification method provided in an embodiment of the present application; Figure 8 A schematic diagram showing a comparison of data between a microfluidic and SPR battery water state identification method with and without a hollow-core fiber installed, provided in an embodiment of the present application; Figure 9 A schematic diagram of the internal structure of a microfluidics and SPR battery water status identification device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0021] To make the purpose, technical solutions, and advantages of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with the specific embodiments of this application and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0022] The embodiments of the present application provide a battery water status identification method, device and medium using microfluidics and SPR, which solve the problems in the prior art of limited contact area between solid optical fiber and liquid, easy destruction of liquid flow distribution after implantation into the flow channel, large liquid flow interference, insufficient integration of microfluidics technology and SPR, and low monitoring sensitivity.

[0023] The technical solutions proposed in the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0024] Figure 1 This is a flow chart of a method for identifying battery water status using microfluidics and SPR, provided in an embodiment of the present application. Figure 1 As shown, the embodiment of the present application provides a method for identifying the water state of a battery using microfluidics and SPR, which specifically includes the following steps: Step 10: Embed the hollow optical fiber into the bipolar plate flow channel of the fuel cell, wherein the cavity of the hollow optical fiber is a microfluidic channel, which is used for the inflow and outflow of liquid inside the fuel cell, and the hollow optical fiber is semi-open.

[0025] In this step, the optical fiber sensor has the advantages of small size, good flexibility, sensitivity to the environment and not susceptible to interference. The SPR (Surface Plasmon Resonance) array has the advantages of high sensitivity, real-time monitoring and versatility. Therefore, the embodiment of the present application uses an SPR optical fiber sensor based on FBG (Fiber Bragg Grating) to perform in-situ monitoring of multiple physical parameters of the fuel cell, and considers embedding the optical fiber sensor into the fuel cell, processing the measured data, and analyzing the state stage of the fuel cell from multiple data by establishing a multi-physical field coupling model, so as to better identify the state of the fuel cell and improve the accuracy and efficiency of diagnosis.

[0026] In the embodiment of the present application, a semi-open cloverleaf hollow fiber (CHF) is used as a sensor. The semi-open cloverleaf hollow fiber structure is as follows: Figure 2 、 Figure 3 As shown, the fiber has four symmetrical pores. During implementation, the pore arrangement can be customized based on actual conditions. The pores within the fiber serve as microfluidic channels, allowing liquid to flow in and out of the fuel cell bipolar plates. Compared to traditional solid-core optical fibers, hollow-core optical fibers significantly increase the contact area between the liquid and the fiber, effectively improving measurement accuracy and making the results more reliable. Furthermore, the liquid between the bipolar plates can flow and transfer through the cavities, which, to a certain extent, reduces the interference of the fiber on the flow characteristics and distribution of the liquid between the plates. Finally, hollow-core optical fibers offer greater real-time performance and accuracy, making them more conducive to real-time monitoring of the battery status within the fuel cell.

[0027] Step 20: Transmit the optical signal to the hollow optical fiber and coat the inner wall of the cavity with a metal layer to stimulate SPR.

[0028] In this step, surface plasmon resonance (SPR) is a physical optical phenomenon typically described using a prism coupling structure. When light is totally reflected between the prism and the metal film surface, an evanescent wave forms. This wave can penetrate the metal and, under certain conditions, resonate with the surface plasmon waves generated by the oscillation of free electrons within the metal. The energy of the evanescent wave is transferred to the surface plasmon waves, absorbing most of the energy of the incident light. SPR utilizes the principle of attenuated total reflection to match the frequency and wave vector of the surface plasmon waves at the metal-dielectric interface with the evanescent waves generated by total reflection in the glass, thus coupling and generating surface plasmon resonance.

[0029] As an optional embodiment, a metal layer is coated on the inner wall of the cavity, and the metal layer is used to excite SPR. Specifically, the process may include: Step 201: coupling an optical signal to a hollow optical fiber via a multimode optical fiber.

[0030] In this step, the spectrometer is connected to the multimode optical fiber, the multimode optical fiber is coupled to the clover-shaped optical fiber, and the optical fiber coupling structure is connected to the spectrometer to excite the optical signal and obtain the detection result signal. The optical signal is effectively transmitted from the multimode optical fiber to the clover-shaped optical fiber through the optical fiber coupling structure to excite the clover-shaped optical fiber SPR. The clover-shaped optical fiber is connected to the fuel cell flow channel to Figure 4 In terms of distribution, PEMFC (Proton Exchange Membrane Fuel Cell) is a proton exchange membrane fuel cell. Multiple monitoring points are set inside the optical fiber to detect the impact of changes in water content at different positions of the fuel cell.

[0031] Step 202: Depositing an adhesion layer on the cavity surface of the hollow optical fiber, the adhesion layer is made of chromium or titanium; Step 203: Coating a metal film on the adhesion layer, and modifying the surface of the metal film with nanoparticles, the metal film is used to excite surface plasmon polaritons, and the metal film is made of gold or silver.

[0032] In this step, the attenuated total reflection coupling structure adopts a fiber-type coupling method, that is, a hollow optical fiber with a diameter of several hundred microns is used as the transmission medium of light. A gold film with a thickness of 50 nanometers and a chromium bottom layer of 2-3 nanometers are coated inside the optical fiber cavity. The purpose of coating the chromium bottom layer is to enhance the adhesion of the gold film to the inner wall of the optical fiber for efficient excitation of SPP (Surface Plasmon Polariton). Gold nanoparticles are modified on the surface of the gold film to improve the SPR sensing performance. When the liquid inside the fuel cell flows through the optical fiber cavity, the evanescent field generated by the light wave transmission interacts with the liquid in the cavity, thereby realizing the sensing of the refractive index of the external environment. The optical fiber cavity is processed using physical vapor deposition (PVD) technology. First, a chromium bottom layer with a thickness of 2-3 nanometers is coated on the surface of the cavity. Then, using this technology, a 50-nanometer gold film is coated on the chromium bottom layer to excite surface plasmon resonance. The relationship between the optical fiber structure, the gold film and the liquid is as follows: Figure 5 shown.

[0033] Step 30: The spectral signal after the interaction between the liquid and the light signal is collected in real time through the microfluidic channel to generate a two-dimensional spectrum diagram, which includes the refractive index, light intensity and resonance wavelength.

[0034] In this step, the experimental data processing module of this scheme is mainly divided into two stages, such as Figure 6The figure shows two parts: the laboratory experiment phase and the application phase. In the laboratory phase, interference was added to obtain spectral images of the fuel cell under normal, flooded, and membrane-dry conditions. Under different liquid contents, the resonance wavelength, refractive index, and light intensity of the spectrum will fluctuate, and these fluctuations will be reflected in the two-dimensional spectrum.

[0035] Step 40: Reduce the dimension of the two-dimensional spectrogram and extract grayscale image features to generate a one-dimensional feature vector.

[0036] As an optional embodiment, the two-dimensional spectrogram is reduced in dimensionality, and grayscale image features are extracted to generate a one-dimensional feature vector. Specifically, the process may include: step 401: converting the two-dimensional spectrogram into a grayscale image and performing pixel compression; step 402: applying a Gabor filter to the compressed image to extract frequency domain features, dividing the image into blocks, calculating the local average value, and generating a low-dimensional feature vector.

[0037] In this step, the dimensionality reduction process is performed on the two-dimensional image, and the two-dimensional image is reduced to a one-dimensional image to obtain the overlapping technology distribution map: the image is first converted to a grayscale image and pixel compression is applied to obtain a 28×28 image, and then a Gabor filter is applied, as shown in the formula:

[0038] The image is then downsampled to reduce the amount of data and extract key features. The image is divided into smaller, non-overlapping blocks, and the average value of each block is calculated to achieve feature dimensionality reduction. The two-dimensional image is converted into a feature vector with 1 row and 51 columns by calculating the GIST features of Torralba (2003).

[0039] Step 50: Input the one-dimensional feature vector into a pre-trained classifier, classify the fuel cell state through supervised and unsupervised hybrid learning, and output a diagnosis result, which includes a normal state, a flooded state, or a membrane dry state.

[0040] In this step, during the application phase, the present solution is applied to actual fuel cell status monitoring, and the obtained image data is input into a classifier model to determine whether the fuel cell is in a normal state, a flooded state, or a membrane dry state.

[0041] As an optional embodiment, the method may also include: when coupling the optical signal through a multimode optical fiber, emitting optical signals of different wavelengths in sequence to obtain the resonance angle offset and the light intensity change, and cross-validating them with the mapping relationship, the mapping relationship being the mapping relationship between the normal state, the flooded state and the membrane dry state and the spectral signal obtained by adding interference during the experimental phase; when the confidence level of the diagnostic results for a preset number of consecutive times is lower than a preset threshold, switching to a redundant sensing mode; correcting the mapping relationship through a self-calibration algorithm, and incorporating the false alarm rate of the fault into the loss function of the classifier.

[0042] In this step, incident light of different wavelengths is emitted sequentially, and the resonance angle offset or light intensity change at each wavelength is analyzed separately. Cross-validation is then performed by combining multi-wavelength signal characteristics with mapping relationships to improve the accuracy of component anomaly judgment. Redundant sensing modes and self-calibration algorithms reduce the false alarm rate, significantly improving the robustness of the system under complex working conditions. The false alarm rate (the probability of misjudging a normal state as flooding) is incorporated into the classifier's loss function, and the model parameters are adjusted through backpropagation to reduce the risk of two types of errors: false alarms and missed alarms.

[0043] As an optional embodiment, a one-dimensional feature vector is input into a pre-trained classifier, and the state of the fuel cell is classified through supervised and unsupervised hybrid learning, and a diagnostic result is output. Specifically, the following may be included: Step 501: extracting a preset number of low-dimensional feature vectors for labeling, and initializing the classifier using the labeled data set; Step 502: based on feature space uncertainty sampling, incrementally learning unlabeled low-dimensional feature vectors with a confidence level lower than a threshold to update the classifier, and evaluating the updated classifier.

[0044] In this step, 30% of the acquired feature data is calibrated, while the remaining data is not calibrated and fed into the established model to classify the fuel cell state represented by the data. The obtained results are further fed back to the model, and the cycle continues. The specific process is as follows: First, the classifier is initialized. The DRB (Discriminative Random Block Model Classifier) is initialized using the labeled training data set:

[0045] Then perform online learning according to the following formula:

[0046] The ASSDRB function is then used to enable the classifier to actively learn from the remaining unlabeled images in an offline mode:

[0047] The updated classifier is then evaluated to test its performance on unlabeled images:

[0048] Through these steps, the ASSDRB classifier can continuously self-update and evolve without retraining from scratch, thereby improving learning efficiency and adapting to new data distributions.

[0049] As an optional embodiment, the method may also include: performing plasma activation treatment on the optical fiber cavity, and setting a vortex generator at the connecting position between the microfluidic channel and the bipolar plate flow channel; using a multi-scale direction combination when extracting frequency domain features using a Gabor filter; and adjusting the temperature of the microfluidic channel to the laboratory temperature value.

[0050] In this step, plasma activation treatment enhances the adhesion of the metal film and prolongs its shedding time in acidic environments. A vortex generator allows the liquid in the bipolar plate flow channel to flow naturally into or out of the fiber cavity, allowing real-time monitoring of water content through SPR technology while maintaining the original flow characteristics of the flow channel. The flow rate and distribution are not significantly disturbed by the fiber implantation, ensuring the normal operation of the fuel cell. A multi-scale directional combination is used: four directions (0°, 45°, 90°, 135°) and three scales (wavelength λ = 8 / 16 / 32 pixels) to generate a 12-dimensional texture feature vector, which is then spliced with the local HOG feature to enhance the spectral image feature expression capability. The reaction chamber temperature is stabilized to a preset value and maintained at this temperature during the detection process to eliminate the interference of temperature fluctuations on the SPR signal.

[0051] As an optional embodiment, after inputting a one-dimensional feature vector into a pre-trained classifier, classifying the state of the fuel cell through supervised and unsupervised hybrid learning, and outputting a diagnosis result, the method may also include: adjusting the liquid flow rate according to the diagnosis result to balance the water content inside the fuel cell; when the diagnosis is a flooded state, increasing the preset discharge rate of the liquid in the microfluidic channel; when the diagnosis is a membrane dry state, reducing the preset discharge rate of the liquid in the microfluidic channel.

[0052] As an optional embodiment, after collecting the spectral signal after the interaction between liquid and light signal in real time through the microfluidic channel, the method may also include: independently extracting time series features of the microfluidic channel and inputting the diagnostic results into the multi-physics field coupling equation; modeling the spatial correlation between channels through the graph convolutional network, predicting the regional diffusion trend within a preset time, and generating preventive control instructions.

[0053] In this step, because the clover-shaped hollow fiber contains four symmetrically distributed microfluidic channels, each channel can independently monitor the liquid flow state at different locations of the fuel cell. There is a spatial coupling relationship between the liquid flow between the channels. The spectral signal of each channel is extracted in the time dimension. Based on the fusion features output by the GCN (Graph Convolutional Network), which contains time series and spatial correlation information, combined with the simulation results of the multi-physics field coupling equation, the diffusion trend of water faults within a preset time in the future is predicted.

[0054] Based on the optical fiber arrangement scheme proposed in the present invention, in order to verify the effect of optical fiber installation on the performance of the bipolar plate, Comsol software was used for simulation. The simulation was performed using a solid optical fiber with a diameter of 130 μm. The installation comparison diagrams are shown in Figures 7 (a) and 7 (b), where Figure 7 (a) shows the flow channel after the optical fiber is installed, and Figure 7 (b) shows the flow channel without the optical fiber installed. The simulation results are shown in Tables 1 and Figure 8 shown.

[0055] Table 1 Simulation results

[0056] According to Table 1 and Figure 8 It can be seen that when a 130-micron diameter optical fiber is installed in the flow channel, the voltage remains essentially unchanged under the same current. Therefore, it can be inferred that the 130-micron optical fiber placement solution is feasible.

[0057] In summary, the present invention combines microfluidics and hollow-core fiber technology to diagnose fuel cell waterlogging. By utilizing the fiber's cavity as a microfluidic channel and coating the inner wall with a gold film, SPR technology is implemented to effectively sense changes in liquid flow within the cavity, thereby distinguishing between flooding, membrane dryness, and normal fuel cell conditions.

[0058] A hollow optical fiber is placed in the bipolar plate flow channel, and the cavity of the optical fiber is used as a channel for the liquid inside the fuel cell to flow out. On the one hand, this method can improve the accuracy of detection and effectively understand the internal working mechanism of the fuel cell. On the other hand, this method can reduce the negative impact on the operation of the fuel cell while ensuring accuracy, and avoid the internal liquid from not being able to flow normally due to the implantation of the optical fiber.

[0059] During the data processing phase, this solution utilizes the ASSDRB classifier, which combines supervised and unsupervised learning techniques, using both labeled and unlabeled data to classify fuel cell states. This method continuously self-updates and evolves without requiring retraining from scratch. This eliminates the need for retraining from scratch, improving learning efficiency, enhancing the model's adaptability to new data, and improving both model stability and the reliability of its results. Furthermore, its ability to self-evolve offers significant advantages in processing real-time data streams.

[0060] The above is an embodiment of the method proposed in this application. Based on the same inventive concept, this application embodiment also provides a microfluidics and SPR battery water state identification device, the structure of which is as follows: Figure 9 shown.

[0061] Figure 9 This is a schematic diagram of the internal structure of a microfluidic and SPR battery water status identification device provided in an embodiment of the present application. Figure 9 As shown, the equipment includes: at least one processor 901; and a memory 902 communicatively connected to the at least one processor; In which, the memory 902 stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor 901 so that the at least one processor 901 can: embed a hollow optical fiber into the bipolar plate flow channel of a fuel cell, wherein the cavity of the hollow optical fiber is a microfluidic channel, the microfluidic channel is used for the inflow and outflow of liquid inside the fuel cell, and the hollow optical fiber is semi-open; transmit the optical signal to the hollow optical fiber, and coat a metal layer on the inner wall of the cavity, and the metal layer is used to excite SPR; through the microfluidic channel, collect the spectral signal after the interaction between the liquid and the optical signal in real time to generate a two-dimensional spectrum graph, which includes refractive index, light intensity and resonance wavelength; reduce the dimension of the two-dimensional spectrum graph, and extract grayscale image features to generate a one-dimensional feature vector; input the one-dimensional feature vector into a pre-trained classifier, classify the state of the fuel cell through supervised and unsupervised hybrid learning, and output a diagnosis result, which includes a normal state, a flooded state or a membrane dry state.

[0062] Some embodiments of the present application provide corresponding Figure 1A non-volatile computer storage medium for microfluidics and SPR battery water state identification stores computer-executable instructions, wherein the computer-executable instructions are configured to: embed a hollow optical fiber into a bipolar plate flow channel of a fuel cell, wherein the cavity of the hollow optical fiber is a microfluidic channel, the microfluidic channel is used for the inflow and outflow of liquid inside the fuel cell, and the hollow optical fiber is semi-open; transmit an optical signal to the hollow optical fiber, and coat a metal layer on the inner wall of the cavity, the metal layer being used to excite SPR; through the microfluidic channel, collect spectral signals after the interaction between the liquid and the optical signal in real time to generate a two-dimensional spectrum graph, the two-dimensional spectrum graph including refractive index, light intensity and resonance wavelength; perform dimensionality reduction on the two-dimensional spectrum graph, extract grayscale image features and generate a one-dimensional feature vector; input the one-dimensional feature vector into a pre-trained classifier, classify the fuel cell state through supervised and unsupervised hybrid learning, and output a diagnosis result, the diagnosis result including normal state, flooded state or membrane dry state.

[0063] The various embodiments in this application are described in a progressive manner. Similar portions between the various embodiments can be referenced to each other. Each embodiment focuses on the differences from the other embodiments. In particular, the IoT device and media embodiments are generally similar to the method embodiments, so their description is relatively simple. For relevant portions, refer to the description of the method embodiments.

[0064] The system and medium provided in the embodiments of the present application correspond one-to-one to the method. Therefore, the system and medium also have similar beneficial technical effects to their corresponding methods. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the system and medium will not be repeated here.

[0065] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0066] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0067] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0068] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0069] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0070] Memory may include non-permanent storage in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0071] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can be implemented using any method or technology to store information. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change RAM (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media such as modulated data signals and carrier waves.

[0072] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0073] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A battery water state identification method based on microfluidics and SPR, characterized in that: The method comprises: Embed a hollow optical fiber into the bipolar plate flow channel of a fuel cell, wherein the cavity of the hollow optical fiber is a microfluidic channel, the microfluidic channel is used for the inflow and outflow of liquid inside the fuel cell, and the hollow optical fiber is semi-open; Transmitting an optical signal to the hollow optical fiber, coating an inner wall of the cavity with a metal layer, wherein the metal layer is used to excite SPR; The spectral signal after the interaction between the liquid and the light signal is collected in real time through the microfluidic channel to generate a two-dimensional spectrum diagram, wherein the two-dimensional spectrum diagram includes a refractive index, light intensity and resonance wavelength; Performing dimensionality reduction on the two-dimensional spectrogram and extracting grayscale image features to generate a one-dimensional feature vector; The one-dimensional feature vector is input into a pre-trained classifier, and the state of the fuel cell is classified through supervised and unsupervised hybrid learning, and a diagnosis result is output. The diagnosis result includes a normal state, a flooded state, or a membrane dry state.

2. The battery water state identification method based on microfluidics and SPR according to claim 1, characterized in that: The coating of the inner wall of the cavity with a metal layer, wherein the metal layer is used to stimulate SPR, specifically comprises: coupling the optical signal to the hollow-core optical fiber via a multimode optical fiber; Depositing an adhesion layer on the cavity surface of the hollow optical fiber, wherein the material of the adhesion layer is chromium or titanium; A metal film is coated on the adhesion layer, and nanoparticles are modified on the surface of the metal film. The metal film is used to excite surface plasmon polaritons, and the material of the metal film is gold or silver.

3. The battery water state identification method based on microfluidics and SPR according to claim 2, characterized in that: The method further comprises: When coupling the optical signal through the multimode optical fiber, optical signals of different wavelengths are emitted in sequence to obtain the resonance angle offset and the light intensity change, which are cross-validated with the mapping relationship obtained by adding interference during the experimental phase and the mapping relationship between the normal state, the flooded state, and the membrane dry state and the spectral signal; When the confidence level of the diagnosis result is lower than a preset threshold for a consecutive preset number of times, switching to a redundant sensing mode; The mapping relationship is corrected through a self-calibration algorithm, and the fault false alarm rate is incorporated into the loss function of the classifier.

4. The battery water state identification method based on microfluidics and SPR according to claim 1, characterized in that: The two-dimensional spectrogram is subjected to dimensionality reduction, and grayscale image features are extracted to generate a one-dimensional feature vector, specifically including: Converting the two-dimensional spectrum image into a grayscale image and performing pixel compression; The Gabor filter is applied to the compressed image to extract frequency domain features, and the image is divided into blocks to calculate the local average value and generate a low-dimensional feature vector.

5. The battery water status identification method based on microfluidics and SPR according to claim 4, characterized in that: Inputting the one-dimensional feature vector into a pre-trained classifier, classifying the fuel cell status through supervised and unsupervised hybrid learning, and outputting a diagnosis result, specifically including: Extracting a preset number of the low-dimensional feature vectors for labeling, and initializing the classifier using the labeled data set; Based on feature space uncertainty sampling, unlabeled low-dimensional feature vectors with confidence levels below a threshold are incrementally learned to update the classifier, and the updated classifier is evaluated.

6. The battery water status identification method based on microfluidics and SPR according to claim 4, characterized in that: The method further comprises: performing plasma activation treatment on the optical fiber cavity, and setting a vortex generator at the communication position between the microfluidic channel and the bipolar plate flow channel; When the Gabor filter extracts frequency domain features, a multi-scale direction combination is adopted; The temperature of the microfluidic channel was adjusted to the laboratory temperature value.

7. The battery water status identification method based on microfluidics and SPR according to claim 1, characterized in that: After inputting the one-dimensional feature vector into a pre-trained classifier, classifying the fuel cell status through supervised and unsupervised hybrid learning, and outputting a diagnosis result, the method further includes: adjusting the liquid flow rate according to the diagnosis result to balance the water content inside the fuel cell; When the flooding state is diagnosed, increasing a preset discharge rate of the liquid in the microfluidic channel; When the membrane is diagnosed as being in a dry state, the liquid in the microfluidic channel is reduced by a preset discharge rate.

8. The battery water status identification method based on microfluidics and SPR according to claim 1, characterized in that: After collecting the spectral signal after the liquid interacts with the optical signal in real time through the microfluidic channel, the method further includes: independently extracting time series features from the microfluidic channel and inputting the diagnostic results into a multi-physics coupling equation; The spatial correlation between channels is modeled through graph convolutional networks, regional diffusion trends within a preset time are predicted, and preventive control instructions are generated.

9. A microfluidics and SPR battery water status identification device, characterized in that: The device comprises: at least one processor; and, a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to: Embed a hollow optical fiber into the bipolar plate flow channel of a fuel cell, wherein the cavity of the hollow optical fiber is a microfluidic channel, the microfluidic channel is used for the inflow and outflow of liquid inside the fuel cell, and the hollow optical fiber is semi-open; Transmitting an optical signal to the hollow optical fiber, coating an inner wall of the cavity with a metal layer, wherein the metal layer is used to excite SPR; The spectral signal after the interaction between the liquid and the light signal is collected in real time through the microfluidic channel to generate a two-dimensional spectrum diagram, wherein the two-dimensional spectrum diagram includes a refractive index, light intensity and resonance wavelength; Performing dimensionality reduction on the two-dimensional spectrogram and extracting grayscale image features to generate a one-dimensional feature vector; The one-dimensional feature vector is input into a pre-trained classifier, and the state of the fuel cell is classified through supervised and unsupervised hybrid learning, and a diagnosis result is output. The diagnosis result includes a normal state, a flooded state, or a membrane dry state.

10. A non-volatile computer storage medium for microfluidics and SPR battery water state identification, storing computer-executable instructions, characterized in that: The computer executable instructions are configured to: Embed a hollow optical fiber into the bipolar plate flow channel of a fuel cell, wherein the cavity of the hollow optical fiber is a microfluidic channel, the microfluidic channel is used for the inflow and outflow of liquid inside the fuel cell, and the hollow optical fiber is semi-open; Transmitting an optical signal to the hollow optical fiber, coating an inner wall of the cavity with a metal layer, wherein the metal layer is used to excite SPR; The spectral signal after the interaction between the liquid and the light signal is collected in real time through the microfluidic channel to generate a two-dimensional spectrum diagram, wherein the two-dimensional spectrum diagram includes a refractive index, light intensity and resonance wavelength; Performing dimensionality reduction on the two-dimensional spectrogram and extracting grayscale image features to generate a one-dimensional feature vector; The one-dimensional feature vector is input into a pre-trained classifier, and the state of the fuel cell is classified through supervised and unsupervised hybrid learning, and a diagnosis result is output. The diagnosis result includes a normal state, a flooded state, or a membrane dry state.

Images